Regulation of virulence related genes by RNA and RNA-interacting proteins in bacteria

(1)

Regulation of virulence-related genes

by RNA and RNA-interacting proteins in bacteria

D I S S E R T A T I O N

zur Erlangung des akademischen Grades

Doctor of Philosophy (Ph.D.)

eingereicht an der

Lebenswissenschaftlichen Fakultät der Humboldt-Universität zu Berlin

von

Andrés Escalera Maurer, MRes in Biochemical Research

Präsidentin der Humboldt-Universität zu Berlin Prof. Dr.-Ing. Dr. Sabine Kunst

Dekan der Lebenswissenschaftlichen Fakultät der Humboldt-Universität zu Berlin

Prof. Dr. Bernhard Grimm

Gutachter/innen

1. Prof. Dr. Kürsad Turgay

2. Prof. Dr. Emmanuelle Charpentier 3. Prof. Dr. Markus Landthaler Tag der mündlichen Prüfung: 27 th May 2019

(2)

to Signe to my family

(3)

Acknowledgements

I would like to thank Manue for allowing me to pursue my interests and crazy ideas surrounded by very special, interesting and nice people. Yan Yan for being a grate example of commitment, passion and dedication. Shi Pey for redefining what hard- work and endurance means. Anaïs for being my unofficial, then official and then unofficial supervisor again, always being interested in my projects and correcting this thesis until the end (even when I asked her to stop). Also for the juggling, always spotting my typos and the writing together (despite the countless file versions). Anne- Laure for showing me that science can be a serious business, for the passionate (and not always focused) discussions and for saving me from some embarrassments in this thesis. Lina for always being there for me, willing to help with anything from making a Halloween costume to organizing my birthday parties. Eric for the long talks about science or life and for his concern in my well-being and survival. Inesita for making sure I didn’t burn down the lab at the beginning (and at the end) and for her interest and effort in becoming more latina. Majda, who was the only person that was not annoyed by me, for being on my side (most of the times) and for being my Cas9 consultant. Franky for being a good friend despite being annoyed by me and for writing the abstract in German (thouh I can’t judge the quality). Laurita for standing my repetitiveness, bad jokes and constant efforts to make her pissed and being a great companion in the office, lunch, dinner, beers, etc. Also for making me feel important and wise by asking me scientific questions of topics she knows more about than me.

Victoria for being an amazing intern and bringing the SagA project back to life, working like crazy and spreading her motivation and joy. Also for making ‘La Cantina’ our second home. Vanessa for her eternal suggestions of new restaurants to try (within one block to her house) and sharing her eternal project with me in a way that made me feel part of it. Katja for making fun of my accent in ‘German’ (than means I know some words).

I also want to thank my TAC members, Kürşad and Markus, for their being tough when I needed it but always beeing interested, constructive and propositive.

I would need another thesis to thank all the people that have helped me and accompanied me during my PhD but I would like to thank all the members of the multiple Charpentier groups that I don’t mention explicitly.

I would like to thank Signe for always supporting and believing in me and for not

(6)

basing her love in the success of my experiments. Also for being involved, up to date and informed about the lab gossip and keeping track of some of my deadlines (that of course I was aware of). To Marianne and Peter who helped with the several movings, if it was not for them I would not have running water in my home. Also for always making me feel welcome in my regenerating visits to Denmark.

My family for their unconditional support and confidence in me. My mum for being always interested in the developments of my PhD, visiting every time she could, integrating in the lab and even attending a Christmas outing. My father for his constant advice, specifically in handling conflicts and lab politics. My brother Manolo for allowing me to see problems from a different perspective or inviting me to escape from them in some other country when I needed it the most. My sister Gena for her interest and involvement in my career planning and encouraging me to continue. My brother Fernando for always being protective, concerned and suppotive of my decisions, despite having different wishes for my future.

(7)

Table of Figures

Figure 1. RNase Y affects expression and processing of the sagA transcript Figure 2. Hemolytic activity is reduced in Δrny compared to the WT strain Figure 3. Truncations of sagA 5′ UTR affect sagA expression levels Figure 4. Truncations in sagA 5′ UTR affect the structure of the RNA Figure 5. Exposure to metabolites affects the sagA 5′ UTR structure Figure 6. Method for validating riboswitches in vitro

Figure 7. Functional analysis of predicted riboswitches in S. pyogenes

Figure 8. Summary of the effects that RNase Y has on sagA at the transcriptional and post-transcriptional levels

Figure 9. FnoCas9 binds to and cleaves its target DNA

Figure 10. FnoCas9 specific binding to its potential RNA targets is not detected Figure 11. Scheme of the immunity and gene regulation mechanisms by the type II-B CRISPR-Cas system of F. novicida

Abbreviations

AHL acyl homoserine lactones asRNAs antisense RNAs

BLP bacterial lipoprotein

Cas CRISPR-associated proteins CDS coding sequence

CRISPR clustered, regularly interspaced palindromic repeats

crRNA CRISPR-RNA

EDTA ethylenediaminetetraacetic acid ES early stationary

FMN flavin mononucleotide

HPLC high-performance liquid chromatography

ML mid-logarithmic

NB Northern blot

ncRNAs non-coding RNAs

nt nucleotides

O/N overnight OD optical density ORF open reading frame

(8)

PAM protospacer adjacent motif PBS phosphate buffer saline pre-crRNA pre-CRISPR RNA

PTS Phosphoenolpyruvate phosphotransferase system qRT-PCR quantitative reverse transcription PCR

QS quorum sensing

RBS ribosomal binding site RNase ribonuclease

RNAseq RNAseq

SAH S-adenosylhomocysteine SAM S-adenosylmethionine

scaRNA small-CRISPR-associated RNA SLO streptolysin O

SLS streptolysin S SP-STK Ser/Thr kinase SP-STP Ser/Thr phosphatase sRNA small RNA

TCS two-component system THY Todd Hewitt broth TL translational

TPP thiamine pyrophosphate tracrRNA trans-activating CRISPR RNA TSA tryptic soy agar

TX transcriptional

WT wild type

YE yeast extract

(9)

Abstract

Bacterial pathogens are constantly regulating the expression of their genes in response to changing environmental conditions and signals from the host. Timely and adequate levels of gene expression are essential for obtaining nutrients and evading the host immune system. The aim of this thesis was to study regulatory mechanisms of virulence-related genes in the bacterial pathogens Francicella novicida and Streptococcus pyogenes.

The focus of chapter one is on the regulation of the important virulence factor streptolysin S (SLS), which is responsible for the hemolytic phenotype of the human pathogen S. pyogenes. First, we investigated the role of the ribonuclease (RNase) Y in the transcriptional and post-transcriptional regulation of sagA, which codes for the precursor of SLS. We found that RNase Y promotes the production of a small RNA (sRNA) from the sagA transcript. However, no role of RNase Y in the regulation of the sagA transcript at the post-transcriptional level was observed. Yet, RNase Y promotes sagA transcription indirectly, affecting the hemolytic activity in a growth phase- dependent manner. Next, we studied the function of sagA 5′ untranslated region (UTR) as a putative cis-acting regulatory RNA. We show that the sagA 5′ UTR contains a secondary structure that may affect the accessibility to the ribosomal binding site (RBS) and that this structure is possibly modulated by direct binding to a ligand.

Moreover, our results indicate that removing fragments of the 5′ UTR has a negative effect on sagA expression, possibly by stabilizing the RBS-blocking structure. While investigating the identity of the putative ligand that affects the sagA 5′ UTR structure, we developed a method for testing the activity of riboswitches. Using this method, we validated three predicted riboswitches in S. pyogenes.

In chapter two, we characterized the mechanism by which F. novicida CRISPR- Cas9 (FnoCas9) represses the expression of bacterial lipoproteins (BLPs), allowing evasion of the host immune system. We show that FnoCas9 is a dual-function protein that, in addition to its canonical DNA nuclease activity, evolved the ability to regulate transcription. In this newly-described mechanism, the non-canonical RNA duplex tracrRNA:scaRNA guides FnoCas9 to the DNA target located downstream of the promoter of the BLP-coding genes (FTN_1103 and FTN_1101), causing transcriptional interference. The endogenous targets contain a protospacer-adjacent motif (PAM) and a sequence that is complementary to scaRNA, promoting FnoCas9 binding. While the

(10)

mechanism is reminiscent of DNA targeting in the canonical immunity function of CRISPR-Cas9, with scaRNA fulfilling a similar function than crRNA, reduced complementarity between scaRNA and the DNA promotes binding but does not allow cleavage. This system can also be engineered to repress other genes, expanding the toolbox of CRISPR applications.

Zusammenfassung

Pathogene Bakterien passen ihre Genexpression konstant an sich verändernde Umweltbedingungen und Einflüsse des Wirtes an. Zeitlich abgestimmte und adäquate Genexpressionslevel sind essentiell für die Nährstoffaufnahme und um einer Immunantwort des Wirtes zu entgehen. Ziel dieser Arbeit war es, die regulatorischen Mechanismen von Virulenz-assoziierten Genen in den Pathogenen Francisella novicida und Streptococcus pyogenes zu untersuchen.

Kapitel eins befasst sich mit der Regulation des wichtigen Virulenzfaktors Streptolysin S (SLS), welcher für den hämolytischen Phänotyp des humanpathogenen Bakteriums S. pyogenes verantwortlich ist. Zunächst untersuchten wir die Funktion der Ribonuklease (RNase) Y während der transkriptionellen und posttranstrikptionellen Regulation des Gens sagA, welches für die Vorstufe von SLS kodiert. RNase Y begünstigte die Produktion einer kleinen RNA (small RNA – sRNA) vom sagA Transkript. Jedoch konnten wir keine Beteiligung der RNase an der posttranskriptionellen Regulierung des sagA Transkripts beobachten. Dennoch förderte RNase Y die Transkription von sagA indirekt, und damit, abhängig von der Wachstumsphase, die hämolytische Aktivität. Weiterhin untersuchten wir die Funktion der 5′-untranslatierten Region (UTR) des sagA Transkripts als ein putatives cis- wirkendes Element. Wir konnten zeigen, dass diese 5′ UTR eine Sekundärstruktur besitzt, die die Zugänglichkeit der ribosomalen Bindungsstelle (RBS) beeinflussen könnte, wobei die Struktur wahrscheinlich durch die Bindung eines Liganden moduliert wird. Außerdem deuten unsere Experimente darauf hin, dass die Deletion einzelner Abschnitte der 5′ UTR einen negativen Effekt auf die Expression von sagA hat, möglicherweise durch die Stabilisierung der RBS-blockierenden Struktur. Um den putativen Liganden zu identifizieren, der die Struktur der 5′ UTR von sagA beeinflusst, haben wir eine Methode entwickelt um die Aktivität von Riboswitches zu analysieren.

(11)

In Kapitel zwei charakterisierten wir den Mechanismus mit dem CRISPR-Cas9 aus F. novicida (FnoCas9) die Expression bakterieller Lipoproteine (BLPs) unterdrückt und damit dem Immunsystem des Wirtes entgeht. Wir konnten zeigen, dass FnoCas9 eine duale Funktion besitzt, die es dem Protein ermöglicht nicht nur DNA zu schneiden (kanonische Funktion), sondern auch Transkriptionsprozesse zu regulieren. Diese erstmals beschriebene Aktivität umfasst die Bindung von FnoCas9 an den nicht- kanonischen RNA-Duplex bestehend aus tracrRNA und scaRNA, wodurch der Protein- RNA Komplex an einen DNA Abschnitt stromabwärts des Promoters zweier BLP- kodierender Gene (FTN_1103 und FTN_1101) bindet und somit eine transkriptionelle Interferenz hervorruft. Diese endogene Bindungsstelle besitzt ein benachbartes Motiv (protospacer-adjacent motif – PAM) und eine scaRNA-komplementäre Sequenz, durch die der FnoCas9-RNA Komplex binden kann. Dieser Mechanismus erinnert an die kanonische DNA-bindende Immunfunktion von CRISPR-Cas9, wobei die scaRNA eine ähnliche Rolle wie die crRNA einnimmt. Jedoch begünstigt die verminderte Komplementarität zwischen scaRNA und der DNA zwar die Bindung, jedoch nicht die Spaltung der DNA. Dieses System kann auch dahingehend verändert werden, um die Expression anderer Gene zu reprimieren und erweitert damit das Repertoire an CRISPR-basierten Anwendungsmöglichkeiten.

(12)

Chapter One:

Regulation of streptolysin S expression by a small RNA and RNase Y in Streptococcus pyogenes

Regulatory RNAs in bacteria

Bacterial non-coding RNAs (ncRNAs) are involved in the regulation of central biological functions such as energy metabolism, quorum sensing, biofilm formation, stress response, adaptation to growth conditions and pathogenesis (Michaux et al., 2014).

Traditional regulatory ncRNAs can be divided in four classes according to their mechanism of action: a) protein-activity modulation, b) antisense c) 5′-encoded regulatory elements (riboswitches and thermosensors) and, d) trans-encoded. Lately, CRISPR (clustered, regularly interspaced palindromic repeats) has emerged as an important new class of ncRNAs that are involved in defense against bacteriophage and plasmid invasion (see chapter two).

Trans-encoded ncRNAs base-pair with an mRNA target and activate or repress translation by diverse mechanisms. When the ncRNA base-pairs near or on the ribosome-binding site (RBS) of the target mRNA, it prevents the ribosome from binding to the mRNA and, therefore, inhibits translation. In other cases, the ncRNA base-pairs upstream of the RBS and promotes translation by inhibiting formation of secondary structures that, in the absence of the ncRNA, block access to the RBS. Additionally, trans-encoded ncRNAs can affect mRNA stability by promoting or inhibiting specific ribonuclease (RNase) activity. mRNA-ncRNA base-pairing can, for example, generate double-stranded (ds)RNA stretches that protect the RNA from single-stranded specific RNases. In other cases, ncRNA binding exposes single-stranded regions that, in the absence of the ncRNA would be double-stranded, allowing single strand(ss)-specific RNases to cleave. Of course, the opposite is also possible, i.e., ssRNA regions of the target mRNA that become ds when bound to the ncRNA, promoting ds-specific- RNases cleavage. Usually these kinds of ncRNAs interact via imperfect complementary sequences with their targets, allowing one ncRNA to have multiple targets (Storz et al., 2011).

Antisense RNAs (asRNAs) are cis-encoded ncRNAs transcribed from the

(13)

opposite strand of their targets and, consequently, fully complementary to them (Georg and Hess, 2018). asRNAs regulate gene expression by affecting transcription, mRNA stability or translation. mRNA stability and translation by cis-encoded sRNAs are regulated by similar mechanisms as those observed for trans-encoded ncRNAs. For asRNA-mediated transcriptional regulation two distinct mechanisms have been proposed: interference and attenuation. Interference means that simultaneous transcription from the sense and antisense strands cause the RNA polymerases to collide, interrupting the process. On the other hand, attenuation occurs when the asRNA causes the formation of a transcriptional terminator in the target mRNA (Sesto et al., 2013). In addition, ncRNAs can regulate protein activity either by acting as cofactors essential for protein activity or by antagonizing or sequestering proteins (Waters and Storz, 2009).

Finally, thermosensors and riboswitches regulate transcription or translation by changing the target RNA structure in response to changes in temperature or presence of a specific molecule, respectively (Ignatov and Johansson, 2017). Riboswitches will be described in more detail in the following section.

Riboswitches

Riboswitches are RNA structures that specifically bind small molecules and modify gene expression. Typically, riboswitches are found in the 5′ UTR of mRNAs but can also be present in ncRNAs such as as and protein-sequestering RNAs (DebRoy et al., 2014; Mellin et al., 2013, 2014). Furthermore, some riboswitch-containing transcripts also act as trans-encoded ncRNAs (Loh et al., 2009). Known riboswitch ligands include ions, cofactors (e.g. vitamins) and modified nucleotides (nt) (such as second messengers). Riboswitches are widely distributed in bacteria and can also be found in some fungi, algae and plants (Barrick and Breaker, 2007; Breaker, 2012). To date, approximately 40 structurally distinct riboswitch classes have been discovered (Lotz and Suess, 2018). Even though each riboswitch class senses a specific ligand, some ligands can be sensed by more than one riboswitch class (Lotz and Suess, 2018).

Riboswitches consist of two elements: a ligand-sensing domain, known as an aptamer, and a regulatory domain, called expression platform. In response to changes in ligand concentration the expression platform undergoes a conformational change that regulates expression of the downstream transcript. Regulatory mechanisms of

(14)

riboswitches include modulation of transcription, translation, transcript stability and processing (Barrick and Breaker, 2007). The expression platform of transcriptional riboswitches can adopt two mutually exclusive conformations: an intrinsic transcriptional terminator that prevents transcription elongation, or an anti-terminator that allows transcription to continue. Similarly, translational riboswitches take two alternative structures that permit or block access to the RBS. Other, less-studied riboswitch mechanisms involve Rho-dependent transcriptional termination and modulation of ribonucleolytic processes, either by self-cleaving ribozymes or ribonucleases (Hollands et al., 2012; Lee et al., 2010; Winkler et al., 2004).

In some cases, multiple regulatory mechanisms can be integrated in one expression platform to give rise to more complex systems, for example by simultaneously regulating translation initiation and cleavage by an RNase (Caron et al., 2012). Ligand binding usually inhibits expression of the adjacent gene although upregulation has also been reported (Mandal and Breaker, 2004; Sudarsan et al., 2008).

Discovery of riboswitches and ligand identification

In order to fulfil their function, aptamers need to specifically recognize their ligand at physiological concentrations (in the pM to mM range, depending on the riboswitch) and discriminate between very similar molecules. This imposes constraints at the level of structure and sequence, which makes the aptamer the most conserved part of riboswitches (Breaker, 2011). This aptamer conservation has been successfully exploited to predict riboswitches (Ames and Breaker, 2010). However, bioinformatics approaches have thus far focused on structures that are widely distributed across species (Barrick et al., 2004; Weinberg et al., 2007). It is therefore likely that riboswitches with narrower distributions have been overlooked. In many cases, the identity of the ligand has been inferred based on the genetic context (Barrick et al., 2004). Yet, finding the ligand in cases where the function of the adjacent gene is unknown can be challenging (Meyer et al., 2011). Moreover, small sequence variations of even a single substitution can alter the specificity of the riboswitch, making it difficult to predict the identity of the ligand even for closely related riboswitches (Weinberg et al., 2017). Therefore, individual riboswitch variants still need to be experimentally

(15)

validated.

Traditionally, a technique called in-line probing has been used to evaluate the binding of ligands to their corresponding riboswitch (Regulski and Breaker, 2008). In- line probing takes advantage of the property of RNA to spontaneously self-cleave in a structure-dependent manner, with single-stranded regions being more prone to degradation (Regulski and Breaker, 2008). Therefore, denaturing-gel electrophoresis analysis of 5′-end labelled RNA after incubation with putative ligands results in a band pattern that provides structural information. Further analysis can even help estimating differences in affinity between closely related molecules (Regulski and Breaker, 2008).

In-line probing has also been used to identify the ligand of a predicted riboswitch from a complex mix of metabolites (Nelson et al., 2013). However, this approach is laborious and, since it renders no functional information, can lead to false positive hits.

Currently, no high-throughput method for identifying ligands of predicted riboswitches exists. In addition to aiding the discovery of endogenous ligands of riboswitches, such a method could be used to identify non-natural ligands of known riboswitches. This knowledge could be harnessed for the development of new antibiotics (Aghdam et al., 2016).

Streptococcus pyogenes

Streptococcus pyogenes is a Gram-positive bacterium that is only known to infect humans. It forms chains of cocci and causes the complete lysis of red-blood cells (beta- hemolysis). Colonization by S. pyogenes can have a wide variety of outcomes, from asymptomatic carriage and mild local colonization in the skin or throat, to deep-tissue and systemic invasions (bacteremia). Pharyngitis (sore throat) is the most frequent disease caused by S. pyogenes. In addition, S. pyogenes is the predominant non-viral cause of pharyngitis (Wessels, 2016). Throat infection, and other streptococcal diseases, can be accompanied by scarlet fever, a skin rash that is likely caused by exposure to streptococcal toxins (Wessels, 2016).

S. pyogenes can also infect different skin layers, causing impetigo or erysipelas when the infection is at the superficial keratin layer and epidermis, respectively (Stevens and Bryant, 2016). Infection of the deeper tissue can lead to more severe diseases such as necrotizing fasciitis, which can have a mortality rate of up to 80 % (Stevens and Bryant, 2016). Superantigens and other virulence factors produced by S. pyogenes may cause an excessive immune response resulting in streptococcal

(16)

toxic shock syndrome and organ failure (Stevens and Bryant, 2016).

Finally, cross-reactivity with the antigens that are present on the surface of S. pyogenes can also lead to post-streptococcal autoimmune sequelae such as acute rheumatic fever leading to rheumatic heart disease or post-streptococcal glomerulonephritis (Cunningham, 2016).

Hemolysins in Streptococcus pyogenes

S. pyogenes secretes multiple virulence factors that help the bacteria obtaining nutrients and in the defense against the immune system of the host (Hynes and Sloan, 2016). Among the most studied virulence factors are the cytolysins, streptolysin S (SLS) and streptolysin O (SLO).

Streptolysin O

SLO is an oxygen-labile pore-forming cytotoxin that is translated as a 69 kDa protein which is activated by a proteolytic cleavage and exported to the extracellular milieu (Hynes and Sloan, 2016). The mature SLS is then inserted in the membrane of host cells in a cholesterol-dependent manner and oligomerizes forming a pore (Hynes and Sloan, 2016). In addition to cholesterol, a galactose-containing receptor is involved in SLO-mediated pore formation in some conditions (Mozola and Caparon, 2015; Shewell et al., 2014). In macrophages, pore formation leads to caspase-dependent apoptosis (Timmer et al., 2009). Consistently, SLO negative mutants are less resistant to killing by macrophages when compared to the isogenic SLO positive strains (Bastiat-Sempe et al., 2014), and are attenuated in virulence (Fontaine et al., 2003; Limbago et al., 2000). Following phagocytosis, SLO prevents acidification, allowing the bacteria to survive (Bastiat-Sempe et al., 2014). SLO also activates neutrophils (Nilsson et al., 2006), promotes inflammation and boosts the immune response (Harder et al., 2009).

SLO is immunogenic and has been proposed as a potential candidate for vaccine development (Chiarot et al., 2013).

Streptolysin S

SLS is an oxygen-stable thiazole/oxazole-modified microcin toxin produced by S. pyogenes and other streptococcal species (Molloy et al., 2011). The genes that are necessary for the production and secretion of SLS are encoded in the nine-gene

(17)

operon sag (Nizet et al., 2000). The first gene of the operon, sagA, encodes the SLS precursor peptide that is modified and exported by the remaining Sag proteins. The genes sagBCD code for a trimeric oxazole/thiazole synthase complex (SagBCD) that modifies SagA conferring it cytolytic activity (Lee et al., 2008). The SagE protease removes the leader of the modified SagA, giving rise to the active SLS toxin (Maxson et al., 2015). Once modified, the active SLS is exported via an ABC transporter formed by the SagGHI proteins (Datta et al., 2005). The remaining Sag protein (SagF) has an unknown function but it is also essential for hemolytic activity (Nizet et al., 2000).

Role in Virulence

The sag operon is conserved across almost all studied strains (Nizet et al., 2000;

Yoshino et al., 2010) suggesting it is important for the survival of S. pyogenes. Indeed, mutant strains that are unable to produce SLS are attenuated in virulence and cause less tissue damage in most murine models of infection, compared to their corresponding isogenic wild type (WT) strain (Betschel et al., 1998; Datta et al., 2005;

Engleberg et al., 2004). The contribution of SLS to virulence varies depending on the model and the studied strain, with some models showing little contribution to survival or pathogenicity (Fontaine et al., 2003; Kinkel and McIver, 2008). In some strains, the relative contribution of SLS to pathogenesis varies in different strains depending on the expression of other factors such as the capsule (Sierig et al., 2003). It was shown that sagA deletion mutant is attenuated in a murine invasive model only when the strain is also unable to produce capsule (Sierig et al., 2003).

Despite the limitations of current infection models for S. pyogenes (Watson et al., 2016), it is now widely accepted that SLS is an important virulence factor for S.

pyogenes (Hynes and Sloan, 2016). However, the specific functions of SLS during infection are less clear. The proposed roles of SLS include defense against the immune systems of the host, dissemination across tissues, and ensuring nutrient availability (Molloy et al., 2011).

The implication of SLS in defense against the immune system of the host is supported by evidence showing that SLS mediates neutrophil and macrophage killing.

Indeed, it was shown that S. pyogenes cytotoxicity on macrophages is mostly mediated by SLS and SLO (Goldmann et al., 2009). It has also been observed that S. pyogenes kills neutrophils in an SLS-dependent manner (Miyoshi-Akiyama et al., 2005). In

(18)

site of infection (Feng et al., 2017; Lin et al., 2009). Consequently, an SLS-deficient mutant was attenuated in virulence and was associated with an increased accumulation of neutrophils compared to the isogenic WT strain (Feng et al., 2017; Lin et al., 2009). In agreement with these results, a recent study has found that SLS activates pain-sensing neurons, which in turn block neutrophils recruitment (Pinho- Ribeiro et al., 2018).

Apart from its role in defense, SLS has been suggested to facilitate the dissemination of S. pyogenes across different tissues. Accordingly, the ability of S.

pyogenes to translocate across epithelial cells in vitro was reduced in a SLS negative mutant compared to the WT (Sumitomo et al., 2011). Interestingly, SLS acts indirectly via the host protease calpain to mediate proteolytic cleavage of intercellular junctions (Sumitomo et al., 2011). A recent study has also linked SLS and SLO with biofilm production in cell cultures and microcolony formation in a mouse model of necrotising fasciitis (Vajjala et al., 2018). This study shows that this is dependent on the ability of the streptolysins to cause endoplasmic reticulum stress and proposes that this promotes biofilm formation, dissemination and proliferation indirectly through the release of unknown signals (Vajjala et al., 2018).

Mode of action

Despite the fact that SLS has been known to lyse cells since 1938 (Molloy et al., 2011), the precise mechanism remains largely unknown. The most detailed biochemical study so far shows that the interaction between SLS and the ion transporter Band 3 mediates lysis of red blood cells by facilitating influx of Cl⁻(Higashi et al., 2016). Furthermore, inhibition of Band 3 activity reduces skin lesion size to similar levels than deleting sagA in a murine model of skin infection (Higashi et al., 2016). However, since the expression of the Band 3 protein is restricted to erythrocyte, the mechanism that mediates SLS-dependent lysis in other cell types is currently unknown.

Regulation

As mentioned above, SLS is an important virulence factor in S. pyogenes. As such, the conditions in which this toxin is produced have been broadly studied. It is important to note that due to inter-strain variability it is impossible to make general conclusions about the role some of these factors have on SLS regulation. Furthermore, even if the

(19)

regulators themselves are conserved, their regulon might vary in different strains and/or conditions. However, because the cues and pathways that affect SLS production are likely related to its function, some general conclusion can be drawn out of this information.

The complete signal transduction pathway linking the input signal to changes in SLS production has not been traced in most cases. Yet, various conditions and cues that affect sagA expression (and some of its regulators) have been discovered. These include nutrient availability (e.g. glucose and nitrogen), growth in blood, saliva or conditioned media and presence of small molecules (such as homoserine lactones, asparagine and SLS autoregulation) (Baruch et al., 2014; Graham et al., 2005; Salim et al., 2007; Saroj et al., 2016, 2017; Shelburne et al., 2010; Steiner and Malke, 2001;

Sundar et al., 2018; Valdes et al., 2018). Some of the factors that modulate SLS production include stand-alone regulators (e.g. Mga, CcpA), two-component systems (e.g CovR/S, Ihk/irr, SptR/S), RNases (i.e. RNases Y, J1, J2, PNPase) and a sRNA (fasX) (Vega et al., 2016).

In spite of the body of knowledge that has accumulated regarding conditions that affect SLS production, there is little information about the mechanisms governing the transcriptional regulation of sagA and even less about the factors affecting SLS production at the post-transcriptional level. The cases where the specific signal that is sensed is known or the regulatory mechanism has been elucidated, are explained in more detail below.

Regulation by small molecules and quorum sensing

Bacteria rely on the production and detection of small molecules in order to sense the presence and abundance of other bacteria in the surrounding environment, a system called quorum sensing (QS).

Sil is a QS system composed of the SilAB two-component system (TCS), the SilDE ABC transporter and the SilCR signalling peptide (Hidalgo-Grass et al., 2002).

Between 12% and 25% of S. pyogenes isolates encode Sil, with some bacteria having incomplete or non-functional systems (Jimenez and Federle, 2014). It has been shown that the pheromone SilCR upregulates sagA expression (Salim et al., 2008).

Interestingly, this effect was observed even in absence of SilAB suggesting the presence of other mechanisms to sense SilCR from other strains or species, even in

(20)

Other QS signaling molecules have been recently implicated in sagA regulation.

Acyl homoserine lactones (AHLs), typically involved in bacterial QS systems, were shown to enter S. pyogenes cells through the ferrichrome transporter FtsABCD and repress sagA expression (Saroj et al., 2017). Regulation of sagA by AHLs is dependent on the QS transcriptional regulator LuxR that was shown to bind sagA promoter (Saroj et al., 2017). However, the exact mechanism mediating this regulation is unclear, as LuxR also seems to bind sagA promoter in the absence of AHLs, at least in vitro. In addition, the inhibitory effect of AHLs was not observed in all the strains studied, suggesting that it might be strain-specific (Saroj et al., 2017). Interestingly, the same study detected an increase in the intracellular iron concentration after addition of AHLs and proposed that inhibition of sagA expression is mediated by iron (Saroj et al., 2017).

Though these hypotheses need further investigation, they are in line with the proposed role of SLS in iron acquisition (Molloy et al., 2011).

In contrast, a previous study found that sagA expression was upregulated in high (1000 µM) compared to low (1 µM) iron concentrations (Salim et al., 2007), which is in agreement with the upregulation of sagA in blood (Graham et al., 2005). The authors proposed that high iron concentrations mimic the environment inside the host phagosome and SLS production allows S. pyogenes to escape (Salim et al., 2007). It is therefore possible that different iron concentrations, or iron signalling under different conditions, have opposing effect.

In addition to AHL, SLS itself has been shown to act as a QS signal via an unknown mechanism (Salim et al., 2007). Conditioned media from WT S. pyogenes but not from a sagA deletion mutant induced sagA expression (Salim et al., 2007). The same effect was observed upon addition of purified SLS to the medium (Salim et al., 2007). This is in contrast to a previous study showing that addition of conditioned media had no effect on sagA expression (Mangold et al., 2004). Therefore, whether SLS acts as a QS molecule or whether it is strain specific remains unclear.

The amino acid asparagine is the only other example where the concentration of a specific molecule is linked to sagA regulation. A study by Baruch and colleagues has found that depletion of asparagine induces expression of the sag operon partly through the TrxRS TCS (Baruch et al., 2014). In addition, SLS and SLO cause endoplasmic reticulum stress, leading to the production of asparagine. Since asparagine promotes S. pyogenes growth in vitro, the authors proposed that one of

(21)

the functions of SLS and SLO is to induce the release of asparagine in order to favour growth (Baruch et al., 2014).

Direct transcriptional regulation

Of all the known transcriptional regulators that affect sagA expression, only two in addition to LuxR (which seems to act in strain-specific manner) have been shown to bind the sagA promoter region: CcpA and CovR.

CcpA is the catabolite control protein that regulates carbohydrate utilization via the phosphoenolpyruvate phosphotransferase system (PTS), which monitors availability of different carbon sources (Deutscher et al., 2006). It was shown that CcpA represses sagA in response to carbon catabolite repression (DebRoy et al., 2016;

Kietzman and Caparon, 2010; Kinkel and McIver, 2008; Shelburne et al., 2008).

However, there is contradictory evidence as to whether this regulation is direct or indirect. While Kietzman and Caparon, 2010 saw no interaction between CcpA and a putative CcpA-binding site upstream of the sagA promoter, others have observed a direct interaction (Kinkel and McIver, 2008; Shelburne et al., 2008). Nonetheless, the regulation of sagA by CcpA is conserved across multiple serotypes (DebRoy et al., 2016) and might explain, at least in part, the repression of sagA expression in the presence of glucose (Sundar et al., 2018; Valdes et al., 2018).

CovR/S is the TCS that controls the expression of several virulence factors.

CovS responds to Mg²⁺ and host antimicrobial peptides (Gordon, 2007; Gryllos et al., 2003, 2008) and phosphorylates the response regulator CovR. Phosphorylated CovR represses sagA expression by binding two sites located in the vicinity of the sagA promoter (Gao et al., 2005; Horstmann et al., 2014). Interestingly, CovR does not require CovS to regulate sagA (Dalton and Scott, 2004; Horstmann et al., 2014), suggesting there are other mechanisms for CovR phosphorylation. Indeed, CovR reversible phosphorylation by the Ser/Thr kinase (SP-STK) and phosphatase (SP- STP) has been shown to affect sagA expression (Agarwal et al., 2011).

Post-transcriptional regulation

In addition to the effect that transcriptional regulators have on sagA expression, there is some evidence, albeit scarcer, indicating that production of SLS might also be regulated at the post-transcriptional level. Four RNases have been shown to affect

(22)

sagA transcript abundance and/or stability (PNPase, RNase Y and RNases J1/J2) (Barnett et al., 2007; Bugrysheva and Scott, 2010; Kang et al., 2010).

In 2007, Barnett and colleagues proposed that two factors are responsible for the increase in sagA transcript abundance at early stationary (ES) growth phase as compared to mid-logarithmic (ML) (Barnett et al., 2007). The first is an increment in promoter activity and the second a stabilization of the sagA transcript (Barnett et al., 2007). They further discovered that the 3′ to 5′ exoribonuclease polynucleotide phosphorylase (PNPase) is involved in decay of the sagA transcript (Barnett et al., 2007). Indeed, while deletion of two other 3′ to 5′ exoRNases (RNase R and YhaM) had no effect in transcript stability, sagA mRNA was 8-fold more stable in a mutant lacking PNPase (Barnett et al., 2007). Though these results do not necessarily mean that PNPase is involved in regulating hemolysis, the difference in sagA stability between the two growth phases suggest that SLS production might be regulated at the transcriptional and post-transcriptional levels.

Other RNases involved in the sagA mRNA degradation are RNases J1 and J2, which are essential in S. pyogenes (Bugrysheva and Scott, 2010). RNase J1 is the only described 5′-to-3′ exoRNase in bacteria and might act as an endoribonuclease in some cases, though this latter activity is still under debate (Durand and Condon, 2018).

RNase J2 is an orthologue of RNase J1 whose activity is less understood but seems to form a complex with RNase J1 (Durand and Condon, 2018). Using conditional mutants of RNases J1 and J2, Bugrysheva and Scott show that the decay of sagA transcript initiates earlier when the expression of the RNases is induced (Bugrysheva and Scott, 2010). These results indicate that RNases J1 and J2 might be involved in the turnover of sagA transcript (Bugrysheva and Scott, 2010).

RNase Y is a single-stranded specific endoRNase that is anchored to the inner side of the membrane (Durand and Condon, 2018). This enzyme is important for the virulence in various Gram-positive bacteria including S. aureus, C. perfringens, and S.

pyogenes (Chen et al., 2013; Kaito et al., 2005; Kang et al., 2010; Khemici et al., 2015;

Marincola et al., 2012; Nagata et al., 2008; Obana et al., 2017). In B. subtilis, RNase Y has an impact on the transcript abundance of most riboswitches (DeLoughery et al., 2018) and other cis-acting RNA structures (Laalami et al., 2013). It has been shown that RNase Y cleaves the SAM-binding riboswitch, preferably in the presence of the ligand (Shahbabian et al., 2009). This suggests that, at least in B. subtilis, RNase Y is

(23)

The biochemical constraints that determine the specificity of RNase Y remain largely unknown. In S. aureus RNase Y was reported to cleave preferably after G in A/U-rich regions (Khemici et al., 2015) and the requirement for a secondary structure downstream of the cleavage site was proposed (Marincola and Wolz, 2017). A recent study from our laboratory showed that a presence of a G is required for RNase Y cleavage but failed to identify any structural requirement (Broglia et al., 2018).

In S. pyogenes, RNase Y was reported to regulate the expression of approximately 29% of the genome, including sagA, which was downregulated in a mutant strain unable to produce RNase Y compared to the WT (Kang et al., 2010).

Although these studies open the possibility that these RNases are involved in the post- transcriptional regulation of sagA, the mechanism and the extent of these effects remain to be investigated.

Results

RNase Y is involved in the processing of sagA transcript

A previous study from our laboratory that used RNA sequencing (RNAseq) and Northern blot analyses to discover novel sRNAs in S. pyogenes showed that sagA mRNA contains a 144 nt-long 5′ UTR that gives rise to a sRNA (Rhun et al., 2016).

Because RNase Y was reported to regulate sagA expression (Kang et al., 2010), it is possible that RNase Y cleaves the sagA 5′ UTR thus regulating sagA expression. Indeed, a ~120 nt-long sRNA was detected in the WT and the Δrny complemented (Δrny::rny) strains but not in the Δrny strain by Northern blot analyses (Figure 1. A-B). Interestingly, even though RNase Y is produced at similar levels throughout the growth phases (Lécrivain, unpublished), this sRNA was observed in ML but not in ES growth phases (Figure 1. B), suggesting that it is produced by a regulated process. Therefore, we first investigated the mechanism by which the sagA transcript was processed and the exact location of the processing site.

Several attempts to determine the exact position of the cleavage by primer extension were unsuccessful (data not shown). It is possible that the downstream fragment produced by RNase Y processing was too unstable to be detected by primer extension. Indeed, cleavage products were not detected with a probe that anneals to the sagA coding sequence (CDS) (Figure 1. C). To overcome these limitations, we

(24)

performed RNA circularization followed by reverse transcription-PCR of the sRNA and sanger sequencing. Yet, transcript ends corresponding to the end of the sRNA were not detected (data not shown). A possible explanation is that the ratio between processed and unprocessed transcripts is low, reducing the probability of detecting the processing site by this technique. In support of this, RNA stability assays showed that the sagA primary transcript was highly stable in the WT and in the Δrny mutant, in both growth phases (Figure 1. D). Because processing of the primary transcript would appear as a reduction of band intensity, this suggested that the rate at which the primary transcript was processed is low.

As an alternative approach to determine the exact location of RNase Y cleavage, we generated deletions and point mutations in the area surrounding the RNase Y putative cleavage site. First, we estimated the approximate location of the 3′

end using the size of the sRNA in the Northern Blot. Then, we constructed a transcriptional reporter fusion to the firefly luciferase gene expressing sagA 5′ UTR under the P23 constitutive promoter (P23-5′ UTR) and introduced substitutions in the RNase Y putative cleavage site (Figure 1. A). Because it was recently reported that RNase Y requires a guanosine (G) adjacent to the cleavage site to be active (Broglia et al., 2018), we introduced G-to-A substitutions in all Gs in this area (Figure 1. A). In addition, we deleted up to 10 nucleotides surrounding the putative cleavage site (Figure 1. A). However, both the substitutions and the deletions failed to inhibit the production of the sRNA (Figure 1. E). Preliminary results suggested that the reporter fusion that contains the WT sequence gives rise to the sRNA in Δrny as well as in the WT (data not shown). Though these results need to be confirmed, this raises the possibility that the sRNA is produced by other mechanisms and that RNase Y affects its production indirectly.

(25)

(26)

Figure 1. RNase Y affects expression and processing of the sagA transcript. A) Schematic representation of the sagA locus (top). Annealing sites of Northern blot probes are indicated with red arrows. Undulated lines represent the transcripts detected by Northern Blot, approximate sizes of transcripts are indicated. The bottom panel shows a close-up of the region surrounding the RBS indicating the sequence and the position of deletions/substitutions introduced in the sagA 5′ UTR.

Numbers indicate coordinates relative to the sagA start codon. B and C) Northern blot analysis of sagA transcript during mid-logarithmic (ML) and early stationary (ES) growth phases in wild-type (WT), RNase Y deletion (Δrny), complementation (Δrny::rny) strain and 5′ UTR sagA deletion (Δ5′UTR) strain. sagA deletion strain (ΔsagA) was used as a control to confirm probe specificity. D) Stability assay of sagA transcript in WT and Δrny grown until ML (left panel) or ES (right panel) growth phases. Transcription was stopped using rifampicin and samples were taken as indicated. Numbers indicate time points in minutes. Approximately 3-fold more RNA was used for Δrny than for WT to compensate for lower initial transcript abundance. E) Northern Blot analysis of ΔsagA transformed with the empty vector (pEC2174), the vector containing the WT sagA 5′ UTR or the 5′ UTR with the mutations described in B, the WT strain was included as a control. The 5S rRNA was used as a loading control for all Northern blots. In E, the size of the 5S RNA band is approximately the same as the band for the sRNA, which interferes with the loading control making it impossible to make any conclusion about the abundance of the sRNA.

However, the presence or absence of the sRNA can be evaluated. The probes used in each experiment is indicated at the bottom. Probe 1 (OLEC3273) anneals at positions −122 to −100 of sagA 5′ UTR (see also A). Probe 2 (OLEC7883) is complementary to the last 26 nt of sagA CDS (see A).

RNase Y regulates sagA mRNA expression

As mentioned above, RNase Y regulates sagA transcript abundance (Kang et al., 2010). In agreement, sagA transcript levels were 4-fold lower in Δrny compared to WT in ML growth phase (Le Rhun, RNAseq differential expression analysis unpublished).

Furthermore, Northern Blot analyses showed that the abundance of sagA primary transcript was lower in Δrny than in the WT or the complemented (Δrny::rny) strain, regardless of the growth phase (Figure 1. B-C). Production of the sRNA should result in lower primary transcript levels, therefore the RNase Y-dependent upregulation of the primary transcript must result from a different process. This indicates that RNase Y has two opposing effects on the abundance of primary sagA transcript (one at the transcriptional and one at the post-transcriptional levels).

In order to investigate the contribution of the post-transcriptional effect, we analyzed the expression of the constitutive P23-5′UTR fusion in the WT and Δrny (Figure 2. A). However, no difference in expression was observed between the two strains in either ML and ES growth phases (Figure 2. A). This is in agreement with the

(27)

fact that the abundance of the processed transcript is low in comparison to the primary transcript (Figure 1. B), suggesting that it was produced at a slow rate.

It is possible that under conditions where the processing rate increases, the post-transcriptional effect is stronger. To investigate the effect that removing sagA 5′

UTR had on transcript abundance, we used the Δ5′ UTR strain, which contained a deletion of the first 122 nt of sagA 5′ UTR in the chromosome of S. pyogenes (leaving 22 nt upstream of the start codon, Figure 1. A). Northern blot analyses showed a lower sagA mRNA abundance in Δ5′ UTR compared to the WT strain, in both growth phases (Figure 1. A). This suggested that a processing event that removes the sagA 5′ UTR from the transcript would have a negative impact on sagA transcript abundance.

To investigate the role of RNase Y in the transcriptional regulation of sagA independently of any post-transcriptional effect, we constructed a fusion of sagA 5′

UTR to the firefly luciferase reporter gene containing the PsagA promoter without the sagA 5′ UTR and tested its expression in S. pyogenes (Figure 2. B, PsagA-fflux).

Unexpectedly, Δrny and WT strains showed similar expression levels for these constructs in both ML and ES growth phases (Figure 2. B). It is possible that the reporter system does not recapitulate the natural regulation due to plasmid copy number or other artefacts.

(28)

Figure 2. Hemolytic activity is reduced in Δrny compared to the WT strain. Activity of reporter fusions containing either the PsagA promoter A) or Δ5′ UTR sagA under the P23 constitutive promoter B) in WT (blue bars) and Δrny (orange bars) cultured until ML or ES growth phases. Luminescence intensity was normalized against the control plasmid (pEC2174, containing the P23 constitutive promoter). Bars show averages for at least three independent biological replicates, error bars represent standard deviations. Streptolysin S (SLS)−dependent hemolytic activity of WT, Δrny and Δrny::rny and Δ5′ UTR sagA in ML (C) and ES (D) growth phases. ΔsagA was used as a control. The average of three independent biological replicates as percentage of the activity of the WT strain is shown. Error bars indicate standard deviation. Independent t−test p values are indicated for relevant comparisons: n.s = p > 0.05, * = p ≤ 0.05, ** = p ≤ 0.01.

(29)

RNase Y deletion affects hemolysis in ML but not in ES growth phase It is clear that the deletion of RNase Y has a negative effect on sagA transcript abundance. In order to test whether RNase Y regulation of sagA transcript has an impact in SagA production, we cloned sagA fused to a Flag tag in a plasmid. However, after several attempts, no signal was detected corresponding to an expressed SagA- Flag in S. pyogenes WT and ΔsagA strains containing this recombinant plasmid by Western Blot (data not shown).

As an indirect approach to detect SagA production, we measured the hemolytic activity of Δrny and WT. As expected, the hemolytic activity of Δrny cultures in ML growth phase was significantly lower when compared to WT or Δrny::rny strains (Figure 2. C). Surprisingly, no difference was observed when the hemolysis assay was carried out using cultures in ES growth phase (Figure 2. D). Because transcript expression is lower in the absence of RNase Y in both growth phases, these results indicate that SLS production is uncoupled from transcript abundance, suggesting that there are additional mechanisms regulating SLS production. In contrast, hemolytic levels were lower in the Δ5′ UTR sagA compared to the WT strain, in both growth phases (Figure 2. C-D), indicating that lower transcript levels are not always compensated by other processes.

Further experiments are needed in order to understand the contribution of RNase Y to the transcriptional and post-transcriptional regulation of sagA expression.

Nevertheless, RNase Y seems to affect the sagA transcript by two potentially independent mechanisms i) inducing the production of a sRNA from the 5′ UTR of the sagA transcript, ii) upregulating sagA transcription through an unknown intermediate factor.

Truncations of sagA 5′ UTR affect sagA expression levels

As shown above, the sagA 5′ UTR is required for WT-levels of sagA expression and SLS production. In order to investigate the regions (and structures) that are important for SagA production, we generated reporter fusions containing various truncations on the 5′ end of the sagA transcript (Figure 3. A). The expression of the truncated fusions was evaluated using the mVenus fluorescent protein as reporter in E. coli (Figure 3. B- D).

(30)

A first set of translational (TL) fusions included, in addition to the 5′ UTR fragment, the first 54 nt of the sagA coding region in frame with the mVenus-coding gene (Figure 3. B). Fusions were named TL−S+s, where S and s are the start and stop coordinates (from the start codon), respectively. The fusion TL−109+54 (lacking the first 35 nt) was expressed at levels similar to those of the full-length fusion (TL−144+54). Interestingly, deleting 28 and 43 additional nt (TL−81+54 and TL−66+54) caused a reduction in fluorescence of approximately 50% and 80%, respectively, when compared to the longer fusions (Figure 3. B). However, removing 39 additional nt (TL−27+54) increased the expression to ~40 % of the TL−81+54 truncation (Figure 3.

B). These results suggest that the integrity of the region downstream of position −109 of the 5′ UTR is important for sagA expression.

In order to test whether the first codons of sagA were involved in the repression of the shorter truncations, we removed the CDS on the above-mentioned fusions.

Similar to first set, the fusion starting at position −66 (TL−66+3) showed lower expression levels (Figure 3. C). However, all truncations were expressed at higher levels (relative to the longest fusion) than constructs containing the sagA CDS fragment. In addition, the shortest fusion (TL−27+3) had similar expression levels that the longest one (Figure 3. C). Together, these results indicate that the absence of the region between −81 and −27 has a negative effect on sagA expression, regardless of the presence or absence of the first 18 codons of sagA CDS.

To determine whether the repression observed in the shorter truncations was due to inhibition of translation, we constructed a set of transcriptional (TX) fusions containing the same regions of sagA 5′ UTR as in the TL fusions. In these fusions, translation of the reporter gene and the fragment of sagA CDS are driven by two independent RBSs. Therefore, any effect can be mostly attributed to changes in transcript levels (Figure 3. D). Similar to the results of the translational fusions, constructs starting at positions −81 and −66 of sagA 5′ UTR had a lower expression than the ones with the full-length 5′ UTR. However, the differences in the expression of the transcriptional fusions are smaller than in the translational fusions and no difference was observed between the fusions starting at positions −81, −66 and −27.

These results indicate that the truncation of 5′ UTR may have an effect on RNA transcription and/or stability.

In order to confirm the importance of the sagA 5′ UTR, we tested the expression

(31)

In agreement with the results from E. coli, the expression of the translational (Figure 3. E) and transcriptional (Figure 3. F) fusions drastically decreased in the fusions lacking 76 nt of the 5′ UTR as compared to the full-length or TX−27+54. Moreover, quantitative reverse-transcription PCR (qRT-PCR) analysis showed that the RNA abundance of TX−66+54 was lower than the full-length and TX−27+54 fusions (Figure 3. G). In contrast to the nearly 90% loss in luminescence (Figure 3. F), RNA abundance of the fusion TX−66+54 was 50% lower than for the full-length fusion (Figure 3. G).

Furthermore, RNA levels of TX−27+54 were similar to the full-length fusion (Figure 3.

G). The discrepancies between the luciferase activity and the RNA levels might indicate that, at least part of the effect, was due to a reduction of translation. Indeed, even if translation of the reporter gene was driven by a separate RBS, the local ribosome concentration might be affected by the proximity of the sagA RBS.

(32)

Figure 3. Truncations of sagA 5′ UTR affect sagA expression levels A) Schematic of reporter fusions containing different fragments of sagA 5′ UTR. sagA CDS is colored red and the 5′ UTR and RBS are colored blue and green, respectively. B and C) Expression of translational reporter fusions under the arabinose inducible promoter (Para) in E. coli containing fragments of sagA 5′ UTR with (B) or without (C) the first 54 nt of sagA CDS. D) Expression of transcriptional fusions in E. coli (similar to B but the translation of mVenus is driven by a separate RBS). Expression of translational (E) or transcriptional (F) reporter constructs containing truncations of sagA 5′ UTR fused to the firefly luciferase gene (Luc) under the P23 constitutive promoter in S. pyogenes. G) Relative RNA abundance analyzed by qRT−PCR in S. pyogenes. In all plots, the labels in the Y−axis indicates the name of the tested fusion. TX and TL are transcriptional and translational fusions, respectively. The numbers indicate the coordinates of the start and end of the sagA sequence counting from the sagA start codon.

Independent t−test p values are indicated for relevant comparisons: n.s = p > 0.05, * = p ≤ 0.05, ** = p ≤ 0.01. The schematic on top of each plot indicates the main features of the tested fusion: fragment of sagA CDS (red), reporter gene and configuration (transcriptional or translational). Bars represent the average of at least three biological replicates relative to the empty reporter vector (fluorescence or luminescence intensity) or the longest fusion (qRT− PCR).

(33)

Structure changes in truncations might inhibit sagA expression As shown above, the sagA 5′ UTR gives rise to an sRNA expressed in a growth phase- dependent manner, suggesting that it is the result of a regulated process. In addition, sagA transcript levels are uncoupled from the hemolytic activity (Figure 2. A and B).

These results could be explained by changes in RNA structure that regulates translation and/or processing.

Interestingly, structure predictions suggest that the RBS is partially buried in a hairpin (Figure 5. A). In addition, this RBS-blocking structure was predicted to occur with a higher probability in the fusion with impaired expression (starting at position −66) than in fusions showing higher expression levels (Figure 4. B). Furthermore, introducing substitutions that were predicted to stabilize the RBS-blocking structure (Figure 4. A and C) drastically reduced the expression of all fusions, independently of the 5′ UTR fragment that was present (Figure 4. D). However, a set of fusions designed to destabilize the structure, with substitutions in predicted base pairs, (Figure 4. C) failed to abrogate the reduction in expression caused by the truncation of the 5′ UTR (Figure 4. D). This is perhaps due to alternative inhibitory structures formed downstream of the RBS (Figure 4. C).

Analyses of fusions containing additional mutations in the 5′ UTR suggested that the region around position −20 was important for sagA expression. Deletion of nucleotides −29 to −20 or −24 to −20, but not deletion of nucleotides −29 to −25, inhibited the luciferase expression to the levels of the fusion that starts at position −27 (Figure 4. A and E). Interestingly, the deletion constructs exhibiting a lower expression were also predicted to have a higher probability of forming an RBS-blocking structure (Figure 4. E). Together, these results suggest that changes in the structure of the 5′

UTR can alter sagA expression.

(34)

Figure 4. Truncations in sagA 5′ UTR affect the structure of the RNA. A) Schematic representation of deletions/substitutions introduced in fusions containing sagA 5′ UTR. Numbers indicate coordinates relative to the sagA start codon. Green and red letters indicate substitutions introduced to generate the ‘open’ and ‘close’ structures, respectively (see panel C). sagA CDS is colored red, the RBS and putative anti−RBS are highlighted in green and yellow, respectively. The UTR is colored blue. B) RNA structure prediction of the full−length 5′ UTR (left panel) or truncations (middle and right panels). C) Structure predictions of the ‘open’ and ‘close’ mutants, expression of these mutants is shown in D. D) Expression of translational fusions containing different fragments of sagA 5′ UTR with substitutions

(35)

that are predicted to generate an Open or Close structure (see C). E) Expression of translational fusions containing deletions of sagA 5′ UTR as indicated (see panel A). F) Structure prediction of the fusions used in E are shown. All structures were predicted using RNAfold web server (Hofacker et al., 1994) and visualized with varna applet. Color−code indicates the probability associated with the position of each nucleotide (red and blue for high and low probabilities, respectively).

Exposure to metabolite mixes affect the sagA 5′ UTR structure

As shown above, structure prediction of the full-length 5′ UTR and truncations indicate that a putative RBS/anti-RBS structure may prevent ribosome binding (Figure 4. A-B).

Stabilization/destabilization of this putative RBS-blocking structure might play a role in regulating sagA translation, potentially explaining the discrepancy between transcript abundance and hemolytic activity (Figure 2. A and B). The presence of a riboswitch in the sagA 5′ UTR could provide a mechanism to regulate accessibility of the RBS and cleavage of the transcript. Indeed, riboswitches control transcript processing and ribosome accessibility, simultaneously (Caron et al., 2012; Shahbabian et al., 2009).

To test the possibility that sagA is under the control of a riboswitch, we used a reporter fusion that contained sagA 5′ UTR and 18 codons of the CDS fused to the gene of mVenus under an inducible promoter. E. coli grown in rich media (EMEM, RPMi or LB) showed a small but reproducible reduction in fluorescence intensity (relative to the vector lacking sagA 5′ UTR) compared with E. coli grown in minimal media (M9, Figure 5. A). This suggested that sagA expression might be inhibited by a molecule present in rich media or a secondary metabolite produced under these conditions. Furthermore, chemical and enzymatic probing of sagA 5′ UTR showed that it adopts a stable secondary structure that could act as a riboswitch or another cis- regulatory element (Figure 5. B-C).

Next, we aimed to determine whether the structure of the 5′ UTR changes in vitro in response to binding of a small molecule. For this, we incubated the in vitro- transcribed sagA 5′ UTR with different concentrations of yeast extract (YE) and assessed whether specific changes in structure occurred. YE was used as a source of metabolites because most known riboswitches sense molecules that are ubiquitous in nature (Lotz and Suess, 2018) and it has been used before successfully for this purpose (Nelson et al., 2013). In order to detect any changes in the structure of the RNA, we used in-line probing (for a description of the technique see introduction, (Regulski and Breaker, 2008). YE caused a concentration-dependent structure change

(36)

on the sagA 5′ UTR (Figure 5. D), suggesting that a component in the YE induced a specific conformational change.

It is known that Mg²⁺is involved in the stabilization of certain RNA structures (Palma et al., 2014). However, a change in structure was still observed upon addition of YE in the absence of Mg²⁺ (Figure 5. E), indicating that Mg²⁺is not strictly necessary for this YE-dependent conformational change to occur. Additionally, the effect of YE was more evident in absence of Mg²⁺. For these reasons, successive experiments were carried out in the absence of Mg²⁺.

One of the proposed functions of SLS-mediated hemolysis is to increase iron availability (Molloy et al., 2011). To determine whether iron was responsible for the observed conformational change, we performed in-line probing in the presence of YE and increasing concentrations of ethylenediaminetetraacetic acid (EDTA), which is a known divalent ion chelator (Figure 5. E). Despite the fact that EDTA inhibited the effect of YE, addition of iron or other divalent cations (in the absence of YE) did not cause specific conformational changes (data not shown, summarized in table 1). Moreover, the effect of YE on the structure of sagA 5′ UTR was reduced when the extract was exposed to high temperatures, indicating that the responsible molecule is heat- sensitive (Figure 5. D). This led us to the conclusion that the structural change is not due to the binding of an ion (Figure 5. F).

In order to identify the putative ligand, we performed subsequent (high- performance liquid chromatography) HPLC/in-line probing cycles (in collaboration with the Chemical Biology Department at HZI, Braunschweig). After each in-line probing experiment, the fraction that caused the conformational change was further separated and re-tested (Figure 5. F). The composition and complexity of the fractions was monitored after each cycle by mass spectrometry and promising candidates were selected for individual testing. Interestingly, a reduced number of fractions for each cycle affected the structure of the RNA, suggesting that the structural change was specific and only caused by a limited number of molecules.

Due to the high complexity of the yeast extract, the identification of potential ligands was challenging. In order to simplify the analysis, a less complex and partially characterized metabolite library from Pseudomonas aeruginosa was used. Both P. aeruginosa cell extract and secreted molecules, in addition to one of the active yeast extract fractions, caused a similar cleavage pattern in a concentration-dependent

(37)

spectrometry analysis, all tested molecules failed to reproduce the pattern caused by yeast and P. aeruginosa extracts (data not shown, see Table 1). This may be due to the fact that the ligand concentration, while still being able to promote RNA structure rearrangements, fell below the limit of detection of mass spectrometry after the HPLC fractionation. A list of the tested compounds can be found in Table 1. These results suggested that the sagA 5′ UTR changes confirmation to a distinct structure in response to the presence of a limited number of metabolites. Future experiments should focus on investigating whether these changes translate into regulation of sagA.

Regulation of virulence related genes by RNA and RNA-interacting proteins in bacteria